RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 03:33:20 +01:00
Code Issues Projects Releases Wiki Activity
64,360 Commits 4 Branches 5 Tags 704 MiB
4a6911275e
Commit Graph

64360 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Chris Lattner
4a6911275e this if can now be an assert. 
llvm-svn: 113173
 2010-09-06 21:25:43 +00:00
Chris Lattner
90ef8a5d5a ; 
llvm-svn: 113172
 2010-09-06 21:23:43 +00:00
Chris Lattner
7b80f3c39c now that the opcode is trivially exposed, start matching instructions 
by doing a binary search over the mnemonic instead of doing a linear
search through all possible instructions.  This implements rdar://7785064

llvm-svn: 113171
 2010-09-06 21:22:45 +00:00
Chris Lattner
e443a993fb emit the match table at global scope instead of within the 
MatchInstructionImpl. This makes it easier to read/understand
MatchInstructionImpl.

llvm-svn: 113170
 2010-09-06 21:08:38 +00:00
Chris Lattner
d323dac9bd special case the mnemonic operand of the instruction in the 
generated matcher, emiting it as a column in the MatchEntry
table instead of forcing it to go through classification and
everything else.  Making it be classified caused tblgen to
produce a ton of one-off classes for each mneumonic.  This
should reduce the size of the generated matcher significantly
while paving the way for future improvements.

llvm-svn: 113169
 2010-09-06 21:01:37 +00:00
Chris Lattner
72d2fb8001 The "ambiguous instructions" check only produces anything with -debug, 
so only do the N^2 loop with debug mode.

llvm-svn: 113168
 2010-09-06 20:21:47 +00:00
Chris Lattner
87234fd589 simplify the hacks around jrcxz. 
llvm-svn: 113167
 2010-09-06 20:10:12 +00:00
Chris Lattner
68f7c5b750 have tblgen detect when an instruction would have matched, but 
failed because a subtarget feature was not enabled.  Use this to
remove a bunch of hacks from the X86AsmParser for rejecting things
like popfl in 64-bit mode.  Previously these hacks weren't needed,
but were important to get a message better than "invalid instruction"
when used in the wrong mode.

This also fixes bugs where pushal would not be rejected correctly in
32-bit mode (just pusha).

llvm-svn: 113166
 2010-09-06 20:08:02 +00:00
Chris Lattner
22bb9cb511 change MatchInstructionImpl to return an enum instead of bool. 
llvm-svn: 113165
 2010-09-06 19:22:17 +00:00
Chris Lattner
faf356777a add note 
llvm-svn: 113164
 2010-09-06 19:14:40 +00:00
Chris Lattner
45a204be76 have AsmMatcherEmitter.cpp produce the hunk of code that gets included 
into the middle of the class, and rework how the different sections of
the generated file are conditionally included for simplicity.

llvm-svn: 113163
 2010-09-06 19:11:01 +00:00
Eric Christopher
fa4de56909 Regenerate. 
llvm-svn: 113160
 2010-09-06 18:47:10 +00:00
Eric Christopher
b26776adb4 Update to 2.9 post-2.8 branch. 
llvm-svn: 113159
 2010-09-06 18:47:00 +00:00
Roman Divacky
f760edb301 Redefine LOOP* instructions from I to Ii8PCRel as they take an i8 argument. 
llvm-svn: 113158
 2010-09-06 18:43:14 +00:00
Chris Lattner
3bd0996c77 random cleanups 
llvm-svn: 113157
 2010-09-06 18:32:06 +00:00
Benjamin Kramer
2447226e73 MCELF: Align symtab, relocation sections and section headers properly. Patch by Krister Wombell. 
llvm-svn: 113155
 2010-09-06 16:11:52 +00:00
Chris Lattner
7cd17a7a4a compute the HasSSE3 bit correctly, patch by Nikolai Saoukh. 
llvm-svn: 113147
 2010-09-06 05:19:44 +00:00
Chris Lattner
6e6a535055 fix PR8067, an over-aggressive assertion in LICM. 
llvm-svn: 113146
 2010-09-06 05:11:24 +00:00
Chris Lattner
1f45ad0aba emit the LLVM intrinsic name -> intrinsic number mapping table with 
StringMatcher instead of a linear sequence of memcmps.

llvm-svn: 113145
 2010-09-06 03:58:45 +00:00
Chris Lattner
cec733bec6 cleanup some of the lifetime/invariant marker stuff, add a big fixme. 
llvm-svn: 113144
 2010-09-06 03:58:04 +00:00
Chris Lattner
8615b37209 allow specifying an indentation level for the string matcher. 
llvm-svn: 113143
 2010-09-06 03:50:59 +00:00
Chris Lattner
8d0ee62652 emit the __builtin -> intrinsic map with StringMatcher instead of a 
copy of a close relative of it.

llvm-svn: 113142
 2010-09-06 03:14:45 +00:00
Chris Lattner
81f696e142 same bug, another place. 
llvm-svn: 113141
 2010-09-06 03:12:27 +00:00
Chris Lattner
6075477174 fix a critical bug where the generated table would say 
"1 strings to match" in a comment, which isn't gramatic.

llvm-svn: 113140
 2010-09-06 03:11:10 +00:00
Chris Lattner
b58e070f0c update cmake 
llvm-svn: 113139
 2010-09-06 02:58:25 +00:00
Chris Lattner
df487455cb factor the snazzy string matcher code that Daniel hates 
out of AsmMatcherEmitter.cpp into its own class.

llvm-svn: 113137
 2010-09-06 02:01:51 +00:00
Chris Lattner
9feb462669 slightly improve the runtime and code size of the Intrinsics info table by not 
comparing the "llvm." prefix in the memcmp, and not storing it in the string literal.

llvm-svn: 113136
 2010-09-06 01:44:44 +00:00
Chris Lattner
7576ffb974 speed up -gvn 3.4% on the testcase in PR7023 
llvm-svn: 113135
 2010-09-06 01:26:29 +00:00
Chris Lattner
4100881939 Teach loop rotate to hoist trivially invariant instructions 
in the duplicated block instead of duplicating them.  

Duplicating them into the end of the loop and the preheader 
means that we got a phi node in the header of the loop, 
which prevented LICM from hoisting them.  GVN would
usually come around later and merge the duplicated 
instructions so we'd get reasonable output... except that
anything dependent on the shoulda-been-hoisted value can't
be hoisted.  In PR5319 (which this fixes), a memory value
didn't get promoted.

llvm-svn: 113134
 2010-09-06 01:10:22 +00:00
Chris Lattner
818045f208 pull a simple method out of LICM into a new 
Loop::hasLoopInvariantOperands method. Remove
a useless and confusing Loop::isLoopInvariant(Instruction)
method, which didn't do what you thought it did.

No functionality change.

llvm-svn: 113133
 2010-09-06 01:05:37 +00:00
Chris Lattner
df0d49314e fit in 80 columns and don't crash on exit, fixes PR8080 
llvm-svn: 113123
 2010-09-05 23:09:30 +00:00
Chris Lattner
c0e5368884 remove some dead code. t2addrmode_imm8s4 is never used in a 
pattern, so there is no need to define a matching function.

llvm-svn: 113122
 2010-09-05 22:51:11 +00:00
Chris Lattner
92429b2601 fix inconsistent formatting. 
llvm-svn: 113121
 2010-09-05 22:43:56 +00:00
Chris Lattner
0bae45de26 cleanups: mark stuff static, only tagdecls should be in anon namespaces. 
llvm-svn: 113120
 2010-09-05 21:25:43 +00:00
Chris Lattner
bd47c48735 cleanups. 
llvm-svn: 113119
 2010-09-05 21:18:45 +00:00
Chris Lattner
091212f130 some random notes. 
llvm-svn: 113118
 2010-09-05 20:49:45 +00:00
Chris Lattner
6c08c3b321 add a comment about where this should eventually move. 
llvm-svn: 113117
 2010-09-05 20:33:40 +00:00
Chris Lattner
908d8e9de2 update this. 
llvm-svn: 113116
 2010-09-05 20:22:09 +00:00
Chris Lattner
e54b9e2332 more cleanups 
llvm-svn: 113115
 2010-09-05 20:13:07 +00:00
Chris Lattner
41d0913a1d Change lower atomic pass to use IntrinsicInst to simplify it a bit. 
llvm-svn: 113114
 2010-09-05 20:10:47 +00:00
Chris Lattner
fc6bf7cb34 eliminate some non-obvious casts. UndefValue isa Constant. 
llvm-svn: 113113
 2010-09-05 20:03:09 +00:00
Chris Lattner
4dccd368f3 fix PR8063, a crash in globalopt in the malloc analysis code. 
llvm-svn: 113109
 2010-09-05 17:20:46 +00:00
Lang Hames
af037cded6 Added initialisers for reduction rule counters. 
llvm-svn: 113108
 2010-09-05 13:42:32 +00:00
Nick Lewycky
d48f0c5718 Fix warning reported by MSVC++ builder. 
llvm-svn: 113106
 2010-09-05 09:11:38 +00:00
Nick Lewycky
97d03432c5 Switch FnSet to containing the ComparableFunction instead of a pointer to one. 
This reduces malloc traffic (yay!) and removes MergeFunctionsEqualityInfo.

llvm-svn: 113105
 2010-09-05 09:00:32 +00:00
Nick Lewycky
b505af7ec4 Fix many bugs when merging weak-strong and weak-weak pairs. We now merge all 
strong functions first to make sure they're the canonical definitions and then
do a second pass looking only for weak functions.

llvm-svn: 113104
 2010-09-05 08:22:49 +00:00
Chris Lattner
684ae57b8e implement rdar://6653118 - fastisel should fold loads where possible. 
Since mem2reg isn't run at -O0, we get a ton of reloads from the stack,
for example, before, this code:

int foo(int x, int y, int z) {
  return x+y+z;
}

used to compile into:

_foo:                                   ## @foo
	subq	$12, %rsp
	movl	%edi, 8(%rsp)
	movl	%esi, 4(%rsp)
	movl	%edx, (%rsp)
	movl	8(%rsp), %edx
	movl	4(%rsp), %esi
	addl	%edx, %esi
	movl	(%rsp), %edx
	addl	%esi, %edx
	movl	%edx, %eax
	addq	$12, %rsp
	ret

Now we produce:

_foo:                                   ## @foo
	subq	$12, %rsp
	movl	%edi, 8(%rsp)
	movl	%esi, 4(%rsp)
	movl	%edx, (%rsp)
	movl	8(%rsp), %edx
	addl	4(%rsp), %edx    ## Folded load
	addl	(%rsp), %edx     ## Folded load
	movl	%edx, %eax
	addq	$12, %rsp
	ret

Fewer instructions and less register use = faster compiles.

llvm-svn: 113102
 2010-09-05 02:18:34 +00:00
Jakob Stoklund Olesen
14f6dc4465 Remove dead code. 
Clobber ranges are no longer used when joining physical registers.
Instead, all aliases are checked for interference.

llvm-svn: 113084
 2010-09-04 21:09:33 +00:00
Chris Lattner
6c6f196725 fix this to work with allocators that have reference type with compilers 
that diagnose invalid references to references.

llvm-svn: 113078
 2010-09-04 18:45:02 +00:00
Chris Lattner
238e7c2fbe dead method. 
llvm-svn: 113077
 2010-09-04 18:19:16 +00:00